Cavium Inc v. Alacritech Inc

1. Case Identification

• Case #: IPR2018-00400
• Patent #: 7,124,205
• Filed: December 27, 2017
• Petitioner(s): Cavium, Inc.
• Patent Owner(s): Alacritech, Inc.
• Challenged Claims: 1, 4-8, 11, 13

2. Patent Overview

• Title: Network Interface Device that Fast-Path Processes Solicited Session Layer Read Commands
• Brief Description: The ’205 patent discloses a network interface device for accelerating data transfers by offloading protocol processing from a host computer. The device uses a "fast-path" to process certain incoming data packets, bypassing the host's network and transport protocol layers, particularly for responses to "solicited read commands" at the session layer.

3. Grounds for Unpatentability

Ground 1: Obviousness over Thia and SMB - Claims 1, 4, 5, 8, 11, and 13 are obvious over Thia in view of SMB.

• Prior Art Relied Upon: Thia (a 1995 article, "A Reduced Operation Protocol Engine (ROPE) for a multiple-layer bypass architecture") and SMB (a 1992 technical standard, "Protocols for X/Open PC Interworking: SMB, Version 2").
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that Thia disclosed the core invention of the ’205 patent: a network interface device (a "ROPE" chip) that performs fast-path processing to bypass a host computer's protocol stack for certain data packets. Thia’s system was explicitly designed for the multi-layer OSI model, which includes session, transport, and network layers. Petitioner contended that SMB disclosed a widely used session-layer protocol for file sharing (Server Message Block) that included "solicited read commands" and was designed to operate within the OSI model. The combination of Thia's fast-path architecture with the specific SMB protocol allegedly taught every limitation of independent claims 1 and 8, including the network interface device, the host protocol stack, and the fast-path processing of a response to a solicited session-layer read command.
  • Motivation to Combine: A Person of Ordinary Skill in the Art (POSITA) would combine Thia and SMB to apply Thia’s theoretical fast-path architecture to a real-world, popular file-sharing protocol. Since Thia’s system required a protocol to be useful, and SMB was a dominant, OSI-compatible protocol for the exact functions described (file sharing), the combination was a natural and logical step. Further motivation existed to improve the performance of SMB communications by offloading a significant portion of the packet processing to Thia's dedicated hardware, a known goal in the art.
  • Expectation of Success: A POSITA would have had a high expectation of success because both references were explicitly designed for or compatible with the OSI protocol model. Thia itself stated its architecture could be adapted to existing systems with "only a small modification," providing an "easy migration path" for protocols like SMB.

Ground 2: Obviousness over Thia, SMB, and Carmichael - Claims 6 and 7 are obvious over Thia in view of SMB and Carmichael.

• Prior Art Relied Upon: Thia, SMB, and Carmichael (Patent 5,894,560).
• Core Argument for this Ground:
  • Prior Art Mapping: This ground built upon the combination of Thia and SMB from Ground 1 to address the additional limitations of dependent claims 6 and 7. These claims required that a response comprising multiple data packets (first and second data) be placed into destination memory in a "substantially contiguous manner" (claim 6) and sequentially (claim 7). Petitioner asserted that Carmichael taught methods for improving data transfer efficiency by using Direct Memory Access (DMA) and physical region descriptor (PRD) tables to transfer multiple, contiguous blocks of data to host memory. Carmichael’s teachings, when applied to the multi-packet SMB responses processed by Thia, allegedly met these specific limitations.
  • Motivation to Combine: A POSITA, having combined Thia and SMB to offload protocol processing, would be further motivated to incorporate Carmichael's teachings to improve the efficiency of the final data transfer step from the network interface device to the host memory. File sharing protocols like SMB routinely transfer files as multiple packets, making efficient DMA and contiguous memory placement critical for overall system performance. Carmichael was directed at improving I/O performance and was compatible with the same operating systems (e.g., UNIX) as SMB, making it a suitable addition to the primary combination.
  • Expectation of Success: The combination would have yielded the predictable result of a fast-path system (Thia/SMB) with a more efficient back-end data transfer mechanism (Carmichael). The integration was straightforward, as Thia already used DMA, and Carmichael simply taught a more advanced and efficient method for managing multi-block DMA transfers to achieve contiguous data placement, a known objective in the field.

4. Arguments Regarding Discretionary Denial

• Petitioner argued that the Board should not exercise its discretion to deny institution under §314(a) or §325(d). An earlier petition filed by Petitioner (IPR2017-01734) with identical grounds and prior art was denied institution solely because Petitioner failed to submit sufficient evidence of the public accessibility of the SMB reference. The Board did not review the substantive merits of the obviousness arguments. This petition was filed to cure that specific evidentiary defect and argued that a merits-based review would not be redundant or an abuse of process, but rather an efficient use of the Board's resources.

5. Relief Requested

• Petitioner requested institution of an inter partes review and cancellation of claims 1, 4-8, 11, and 13 of the ’205 patent as unpatentable.